Information Technology Reference
In-Depth Information
recordings are particularly suited for studying spatiotemporally coherent and locally
synchronized collective neural dynamics. There is a limit to how much current den-
sity a patch of cortex can support [62], thus large amplitude fields/potentials entail
distributed synchronized oscillations.
From a practical standpoint, signals may be contaminated by EM artifacts orig-
inating from the heart, muscles, eyes, and the environment. These sources of per-
turbation to the estimation of neural currents can be attenuated and/or corrected
using appropriate denoising techniques. These include noise cancelation using fil-
ters in the temporal, spatial, and frequency domains (see Section 12.3 and [6] for a
review).
8.2 Measurement Modalities
All the EM recording techniques share the important benefit of high sampling rates
during acquisition (up to 5 KHz on several hundreds of channels). However, they
measure different, yet closely related physical quantities at different spatial scales.
In principle, the inverse modeling methods described here can be applied to data
acquired using MEG, EEG, ECoG, and combinations of these measurement modal-
ities. A prerequisite is the modeling of source currents, tissue geometry and conduc-
tivity, and sensor technology. This has yield an abundant literature in the domain of
forward modeling techniques, which is reviewed in Section 8.4.4.
8.2.1 Magnetoencephalography (MEG)
In MEG, an array of sensors is used to noninvasively measure components of the
magnetic vector field surrounding the head [31, 97]. The magnetic fields generated
by neurons are extremely weak and range by about a billion times smaller than
the Earth's static magnetic field. This low signal-to-noise ratio (SNR) challenged
the early development of MEG technology. The first magnetoencephalogram was
recorded with a single heavily wounded coil [12]. Not long after, the superconduct-
ing quantum interference device (SQUID) was invented [103]. This extremely sen-
sitive magnetometer (consisting of a superconducting loop with one or two Joseph-
son junctions), coupled to a pickup coil via a flux transformer, allowed for the first
low-noise MEG recordings by the early 1970s [13]. For a thorough overview of
SQUID electronics and modern integrated thin-film magnetometers and gradiome-
ters, see [31]. Importantly, to dramatically increase the SNR, MEG measurements
are acquired inside a magnetically shielded room (MSR). Current state-of-the-art
systems include a large number of sensors (
>
300), organized as a helmet-array of
magnetometers and/or gradiometers (planar or axial) that can measure spatial gra-
dients of the magnetic field. This latter arrangement has been demonstrated to be
beneficial to the SNR by attenuating environmental perturbations. Distant reference
